3 research outputs found
Microwave Imaging Technique for Detection of Multiple Line Cracks in Concrete Material
Full text linkNowadays, many concrete structures are exposed to higher loads than they are designed for due to the increase in human population and uncertain environmental conditions. This leads to a faster deterioration of the structure for example formation of cracks. Cracks provide significant signs for the health and residual strength of a civil structure. Though they appear at times in form of lines, they cannot be easily identified or detected by traditional methods and techniques. For example, manual inspection is too costly in terms of time and effort; meanwhile, other non-destructive techniques are bounded by each unique weaknesses. This research proposes a microwave imaging technique with ultra-wideband (UWB) signal in detecting multiple line cracks. Various crack scenarios were first simulated using Finite-Difference Time-Domain (FDTD) to see the performance of the proposed technique. Delay-And-Sum algorithm is used for image reconstruction. This technique was able to detect single, double, three and multiple line cracks on single brick of size 2 mm accurately. Notwithstanding, more than three cracks on a single brick could not be detected as they appear as a clutter. In conclusion, the proposed technique is useful for crack detection in building and man-made structures
Development of a Fall Detection System Based on Neural Network Featuring IoT-Technology
Full text linkAccidental falls are considered a major cause of accidents that could lead to serious injuries, paralysis, psychological damage, and even deaths, especially for the elderly. Therefore in this project, a neural network-based fall detection system that could automatically detect a fall event is proposed. The system is enhanced with Internet-of-Things (IoT) features that could reduce the response time and efficiently improve the prognosis of fall victims. A 10 Degree of Freedom (DOF) Inertial Measurement Unit (IMU) module is connected to an Intel Edison with Mini Breakout board and mounted on a wearable waist-worn device to continuously record body movements. A backpropagation neural network algorithm has been developed to accurately distinguish falls from different postural transitions during activities of daily living (ADL). A body temperature and heart-pulse monitoring device were developed for this system to provide the medical personnel additional information on the body condition of the fall victim. Using the latest IoT-technology, the system can be connected to the internet and provides a continuous and real-time monitoring capability. Once a fall accident happens, the system will be automatically triggered. This will activate an Android App through the Wi-Fi network that will then send an emergency SMS with the actual location and body conditions of the victim to a recipient. A series of falls and ADL simulations were performed by a group of subjects to test and validate the performance of the system. The experiment results showed that the proposed system could obtain a sensitivity of 95.5%, specificity of 96.4%, and accuracy of 96.3%
Spectrogram Based Window Selection For The Detection Of Voltage Variation
Full text linkThis paper presents the application of spectrogram with K-nearest neighbors (KNN) and Support Vector Machine (SVM) for window selection and voltage variation classification. The voltage variation signals such as voltage sag, swell and interruption are simulated in Matlab and analyzed in spectrogram with different windows which are 256, 512 and 1024. The variations analyzed by spectrogram are displayed in time-frequency representation (TFR) and voltage per unit (PU) graphs. The parameters are calculated from the TFR obtained and be used as inputs for KNN and SVM classifiers. The signals obtained are then added with noise (0SNR and 20SNR) and used in classification. The tested data contain voltage variation signals obtained using the mathematical models simulated in Matlab and the signals added with noise. Classification accuracy of each window by each classifier is obtained and compared along with the TFR and voltage PU graphs to select the best window to be used to analyze the best window to be used to analyze the voltage variation signals in spectrogram. The results showed window 1024 is more suitable to be used
